ENVIRONMENTAL AUDITING
An Integrated Environmental Assessment of the US
Mid-Atlantic Region1
J. D. WICKHAM*
US Environmental Protection Agency (MD-56)
Research Triangle Park, North Carolina 27711, USA
K. B. JONES
US Environmental Protection Agency
Las Vegas, Nevada 89119, USA
K. H. RIITTERS
Biological Resources Division
US Geological Survey, Raleigh, North Carolina 27695, USA
R. V. O’NEILL
Environmental Sciences Division
Oak Ridge National Laboratory, Oak Ridge, Tennessee
37831, USA
R. D. TANKERSLEY
Department of Geography
University of Tennessee, Knoxville, Tennessee 37996, USA
E. R. SMITH
US Environmental Protection Agency
Research Triangle Park, North Carolina 27711, USA
Environmental assessments, mandated by the National Environmental Policy Act (NEPA), require
examination of cumulative impacts (Shoemaker 1994).
Cumulative impact assessment has tended to concentrate on a specific proposed activity (see Fabos 1985) or
a potentially threatened resource (Lee and Gosselink 1988). Many of today’s environmental problems occur at larger spatial scales that span regions
and even continents (Hunsaker and others 1990),
including habitat fragmentation and land-use change,
modification of streams, and air pollution. As a result, they have the potential to occur at the same place
1This
research was funded by the US Environmental Protection Agency
(EPA) through its Office of Research and Development. The paper has
been reviewed by EPA and approved for publication. Funding for R. V.
O’Neill was through interagency agreement DW89936104-01-0 with
Oak Ridge National Laboratory. Mention of trade names or commercial products does not constitute endorsement or recommendation for
use by EPA.
KEY WORDS: Cluster analysis; Cumulative impact; Geographic information systems; Landscape ecology; Remote sensing
*Author to whom correspondence should be addressed.
Environmental Management Vol. 24, No. 4, pp. 553–560
A. C. NEALE
D. J. CHALOUD
US Environmental Protection Agency
Las Vegas, Nevada 89119, USA
ABSTRACT / Many of today’s environmental problems are
regional in scope and their effects overlap and interact. We
developed a simple method to provide an integrated assessment of environmental conditions and estimate cumulative impacts across a large region, by combining data on
land-cover, population, roads, streams, air pollution, and
topography. The integrated assessment technique identified
nine distinct groups of watersheds. Relative cumulative impact scores were highest around major urban centers, but
there was not a simple or predictable spatial pattern overall.
We also point out the potential applications of this approach
that include distinguishing between areas in relatively poor
versus good condition, identifying areas that may be more
vulnerable to future environmental degradation, and identifying areas for restoration.
and at the same time. Because of the possibility of
cooccurrence, their impacts can be cumulative and new
approaches are required to provide an integrated assessment.
Like the terms stress, disturbance, and perturbation
(see Rykiel 1985), the term impact has been used in
many contexts in the ecological literature. It has been
used as a synonym for effect when a causal link can be
established (Beanlands and others 1986). The term
impact comes from the Latin word impactus, the past
participle of impingere, which means to strike against. In
an ecological context, the Latin derivative is equivalent
to the term disturbance (i.e., a physical force, agent, or
process; sensu Rykiel 1985). We use the term here
as an equivalent of disturbance, not as a synonym for
effect.
The recognition of cumulative impacts (Odum
1982) stimulated considerable research in the United
States and Canada (Environmental Management
1988, Irwin and Rodes 1992, Shoemaker 1994). This
research produced a categorization of the types of
cumulative impacts (Beanlands and others 1986):
(1) multiple disturbances of a single kind overlapping
r 1999 Springer-Verlag New York Inc.
J. D. Wickham and others
554
Table 1.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
Indicators of regional ecological conditions
Population density (1990)
Population change (1970–1990)
Road density
Percent of watershed in anthropogenic land cover
Percent of watershed streamlength with roads within 30 m
Average atmospheric wet NO3 deposition (1987 and 1993)
Average atmospheric wet SO4 deposition (1987 and 1993)
Percent of watershed with cropland on slopes ⬎3%
Percent of watershed with cropland and pasture on slopes
⬎3%
Number of impoundments per 1000 km of stream
Percent of watershed that is forested
Percent of watershed streamlength with agriculture within
30 m
Estimated N load in streams
Estimated P load in streams
Percent of watershed streamlength with forest within 30 m
Soil loss (estimated from USLE)
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
Forest density (1 ha window)
Forest density (65 ha window)
Forest density (600 ha window)
Proportion of watershed that supports forest at three
scales (17, 18, 19)
Largest forest patch (expressed as proportion of
watershed area)
Density of forest edges in 1 ha
Density of forest edges in 65-ha window
Density of forest edges in 600-ha window
Percent of watershed showing vegetation loss
Percent of watershed showing vegetation gain
Proportion of watershed showing vegetation change
Proportion of watershed showing vegetation loss on slopes
⬎3%
Vegetation loss in first-order stream region
Vegetation gain in first-order stream region
Vegetation change in first-order stream region
aPopulation
and population change estimates are based on data from US Department of Census. Per watershed estimates of density were based on
models re-assigning county estimates to square kilometer cells based on road density (Xie 1995).
bRoads
and streams are from USGS 1:100,000-scale Digital Line Graph Data (USGS 1989).
cLand
cover component of indicators 4, 8, 9, and 11 through 24 were derived from ca. 1990 Landsat TM-based land cover data, from
Multi-Resolution Landscape Characteristics Consortium (MRLC).
dAtmospheric
wet deposition data were from USEPA and surface maps of these data were based on modeling at Pennsylvania State University. Values
are watershed means in kg/ha/yr ⴱ 100.
eSlope
estimates are from USGS 1:250,000-scale Digital Elevation Model (DEM) Data (USGS 1993).
fSoil
loss is expressed as the proportion of the watershed where estimated soil loss is greater than 1 ton/acre/yr, based on the USLE model
(Wischmeier and Smith 1978).
gNitrogen
and phosphorus loads (kg/ha/yr) in streams were estimated using simple screening models (Rechow et al. 1980, Soranno et al. 1996).
hMultiple
window indicators (17–20 and 22–24) simulate habitat suitability for small, medium, and large wildlife (Riitters et al. 1997).
iIndicators 25–31 are based on multitemporal comparison of Normalized Differenced Vegetation Indices (NDVI) compiled from ca. 1975 and 1990
Landsat MSS data (North American Landscape Characterization Program). Indicators 25–28 are expressed as a proportion of the watershed area.
Indicators 29–31 are expressed as a proportion of the first-order stream region in the watershed. The first-order stream region is that closest to
first-order streams.
in time, (2) multiple disturbances of one or more
types overlapping in space, (3) indirect effects,
and (4) accumulation of small, apparently insignificant disturbances that result in a significant impact in total.
The purpose of this paper is to present a method and
results for an integrated assessment of multiple disturbances that overlap in space (item 2 above). The
research uses geographic data covering the five midAtlantic states of Pennsylvania, Maryland, Delaware,
Virginia, and West Virginia. A fundamental difference
between traditional and cumulative impact analysis is
that the latter takes a broader spatial and temporal view
(Preston and Bedford 1988). A broader spatial (and
temporal) view requires use of data from satellites and
other geographic sources.
Methods
Monitoring and integrated assessment at a regional
scale has become feasible because of three recent
developments: (1) the availability of remote imagery,
(2) geographic information systems (GIS), and (3)
advances in landscape ecology. Landscape ecology
makes it possible to relate land-cover to a number of
ecological variables (Riitters and others 1995, 1996,
Jones and others 1997, Wickham and others 1997,
O’Neill and others 1997). We analyzed 31 indicators
(Table 1) on a watershed-by-watershed basis, using US
Geological Survey (USGS) hydrologic unit maps (USGS
1982).
Correlation and minimum-Euclidean-distance-tomean cluster analysis were used to identify groups of
watersheds with similar data values, and canonical
discriminant analysis (CDA) was used to check
the distinctness of the clusters. CDA is conceptually
similar to principal component analysis in that the
input variables can be summarized as a reduced set and
the clusters plotted as a function of the canonical
scores.
Integrated Environmental Assessment
Cluster analysis is an exploratory method that does
not have well-developed validation techniques (Aldenderfer and Blashfield 1984). Several iterations of
the cluster analysis were performed, using different
combinations of variables and transformations. The
clustering output that provided the most distinct grouping in canonical space was taken as the solution. Five of
the nine clusters from the final output were largely
consistent across all iterations.
The final subset of indicators used included items 1,
2, 3, 5, 7, 9, 10, 15, and 20 (Table 1). These variables
are related to several different aspects of environmental condition, including population (1, 2, 3), habitat (3, 15, 20), water quality (3, 5, 9, 15), pollution
(7, 9, 15), and potential impairment of hydrologic function (10). Pair-wise correlation coefficients
were less than ⫾0.5 except in two cases. Population
density (1) and road density (3) had an r value of
⫹0.75. Both were included to weight population
more heavily, because many landscape variables
(e.g., 9, 15, 20) are predictable when population is
high even though the variables are not correlated
across all watersheds. Agriculture on steep slopes (9)
and proportion of streamlength with adjacent
forests (15) were moderately inversely correlated
with each other (r ⫽ ⫺0.67). Both were included because they measure different aspects of the environment. The former is a measure of potential soil erosion
and water pollution (Wischmeier and Smith 1978,
Renard and others 1997), while the latter is a measure
of potential habitat and nutrient filtering capacity
(Wharton and others 1982, Peterjohn and Correll
1984).
The final subset of indicators was transformed
prior to clustering. Square root transformations
were used for variables that had observations with
zero values, (9, 10, 20). Logit transformations
(Evans and Jones 1981) were used for proportions
(5, 15), and logarithmic transformations were used for
the remaining variables. Data values ranged between
⫺10 and 10 after transformation. Other transformations, including standardizing to a mean of zero and
variance of one, did not improve separation in canonical space.
We followed the recommended procedure for identifying clusters, (SAS 1989, p. 883). An initial clustering
was run to establish seed values. The initial clusters
selected as seed values had at least five observations,
and the maximum distance of an observation
to the cluster mean was less than the centroid-tocentroid distance to the nearest cluster, except in one
case. For the one exception, the between-cluster distance (2.37) was slightly less than the within-cluster
555
distance (2.53). A second iteration was run to assign
watersheds to clusters based on the seed values established.
The cluster means table provides an integrated
environmental assessment and a way to identify
relative cumulative impacts. It is an n x m matrix, where
the rows (n) represent the clusters and the columns (m)
represent the cluster’s mean value for each of the
indicators. Thus, the mean values for each of the
indicators, taken together, constitute an integrated
environmental assessment of watershed groups. Ranking for relative cumulative impact was done by reading
down the columns to identify the cluster means representing the poorest conditions. After identifying the
poorest values in each column, relative cumulative
impact was determined by reading across the rows and
tallying the number of poor values. Clusters with a
higher total of poor values have a greater relative
cumulative impact score than clusters with a lower total
of poor values.
Results
Cluster analysis identified nine groups of watersheds (Figure 1). A graph of the clusters as a function of the first and second canonical scores shows
that clusters 1, 2, 3, 8, and 9 occupy distinct regions of
the X–Y space, and clusters 5 and 6 are well separated
from clusters 4 and 7 (Figure 2). Distinct separation of
cluster 5 from 6 and cluster 4 from 7 was evident on
graphs of the first and third and first and fourth
canonical scores, respectively. The first four canonical
variables explained about 96% of the variance in the
data.
Cluster means are shown in Table 2. The
rows represent the clusters and columns represent
the indicators. The mean indicator scores for each
cluster (columns) were separated into three, equalfrequency groups (best, middle, worst). This provided a
simple and straightforward way to group watersheds based on indicator scores. The three worst
values (highlighted in bold) were tallied across
each row to get a relative cumulative impact score
for each cluster. Clusters with higher totals of poor
values have greater relative cumulative impact scores
than clusters with lower totals of poor values. Relative
cumulative impact scores ranged from zero to
six. Watershed groups are described in rank order in
Table 3.
There is no simple spatial pattern to the rankings,
such as scores increasing from the less populated west to
the more populated east (Figure 1). Only three of the
556
J. D. Wickham and others
Figure 1. Watershed groups from cluster analysis (see Table 2 for interpretation).
nine groups are spatially contiguous. For example,
clusters 3 and 8 are intermingled throughout the
Appalachian Mountain region, but have relative cumulative impact scores of 4 and 1, respectively. On average,
watersheds in cluster 3 have a greater proportion of
their streamlength near roads, a greater proportion of
agriculture on steep slopes, less forest adjacent to
streams, and less forest in large contiguous blocks.
Integrated Environmental Assessment
557
Figure 2. Watershed groups as a
function of first and second canonical scores.
Discussion
Many of the indicators used to develop this integrated assessment can be related to aspects of the
environment considered important by society (Westman 1977, O’Neill and others 1994, Kepner and others
1995). Indicators C–H in Table 2 provide broad-scale
measurements related to the amount and quality of
upland habitat and ecological condition of streams.
Riparian forests help to prevent pollution (Peterjohn
and Correll 1984) and provide habitat (Wharton and
others 1982). Agriculture on steep slopes increases the
potential for soil erosion and delivery of sediment to
streams (Wischmeier and Smith 1978, Renard and
others 1997). Construction of dams often leads to
geomorphic changes, which in turn lead to ecological
changes in the stream ecosystem (Ligon and others
1995). Loss of forest removes habitat (Lynch and
Whigham 1984), a source of atmospheric carbon sequestration (Wessman 1992), and changes the local climate
(Pielke and Avissar 1990).
Mixing indicators of human influence and broadscale ecological condition to derive relative cumulative
impact scores does not necessarily assume that increasing human influence results in degraded ecological
condition. There is not a one-to-one correspondence in
rank orders of relative cumulative impact scores and
population density (see note to Table 2). For example,
the population density of cluster 1 is more than twice
J. D. Wickham and others
558
Table 3. Description of watershed groups based
on ranking by relative cumulative impact score
Table 2. Cluster means and relative cumulative
impact score for watershed groupsa
Cluster
A
B
1
2
3
4
5
6
7
8
9
21.5
31.6
7.5
⫺3.5
10.9
⫺2.6
66.7
10.1
⫺4.0
2.5
1.7
1.9
1.7
2.4
4.1
1.5
1.3
1.1
C
D
E
F
G
6.3 8.9 84.5 20.2 2272
2.5 0.3 82.5 0.6 2435
6.9 15.2 72.6 16.3 2468
1.7 6.0 93.6 5.8 1877
4.8 20.2 70.8 1.3 2825
7.1 7.2 73.6 0.8 2607
1.9 1.1 88.6 6.7 2056
5.5 9.7 83.8 37.5 2377
9.9 3.3 90.0 69.0 2290
H
R.C.I.
42.0
0.2
8.7
12.0
2.9
15.7
19.3
4.4
4.3
3
2
4
0
5
6
2
1
1
*Population density was not used for relative cumulative impact
scoring because county-based population densities were mapped to
watersheds using road density. Average population densities for clusters 1–9 are respectively: 175, 111, 88, 28, 325, 982, 78, 38, and 29
persons per square kilometer. A ⫽ population change (1970–1990);
B ⫽ road density; C ⫽ proportion of watershed streamlength that had
roads within 30 m; D ⫽ proportion of watershed with cropland and
pasture on slopes ⬎3%; E ⫽ proportion of watershed streamlength
with adjacent forest; F ⫽ proportion of watershed supporting forest at
three scales; G ⫽ average atmospheric sulfate wet deposition (1987 and
1993); H ⫽ number of impoundments per 1000 stream kilometers;
R.C.I. ⫽ relative cumulative impact.
that of cluster 3, but cluster 1 has better scores for most
of the broad-scale measurements of ecological condition. Comparison of indicators A, B, and population
density with indicators C–H help to show that the
relationship between human occupancy of the landscape and ecological condition is not necessarily straightforward.
There are several applications of the cluster means
table that are related to environmental analysis and
management. First, the cluster means table permits easy
identification of the watersheds that are least and most
impacted, based on the data used. Second, these data
can be used to find watersheds that may be more
vulnerable to future environmental degradation. Watersheds in cluster 1 have forests in relatively large,
contiguous blocks, but they also have a fairly high
human population density that is increasing rapidly.
These watersheds may be more vulnerable to future
forest fragmentation than watersheds in other groups.
Third, the information in the cluster means table can
also be used to guide (and test) ecological restoration
(Allen and Hoekstra 1987). For example, watersheds in
cluster 4 have a fairly high percentage of forest cover
(about 70%), but a low proportion of their watershed
area in larger blocks of forest (cluster mean score of 5.8
for variable F in Table 2). Since the abundance of forest
interior birds in this region has been shown to be
positively correlated with forest size and/or negatively
correlated with forest isolation (Lynch and Whigham
1984), reforestation so that large blocks are formed
Rank 1 (Cluster 4): None of the cluster means for this group
are among the poorest values. Its cumulative impact score
is 0.
Rank 2 (Cluster 9): Watersheds in this group have the highest
amounts of forest and riparian cover, and impacts from
roads, agriculture, and impoundments are low. This group
does have the poorest score for roads adjacent to streams,
and, hence, a cumulative impact score of 1. However, road
density is low. The high value for adjacency to streams may
be due to the group’s occurrence on the Appalachian
Plateau where there is a higher proportion of land on steep
slopes. There may be a tendency for roads to follow streams
along the narrow valleys.
Rank 2 (Cluster 8): The relative cumulative impact score for
this group is also 1. The group has a poor score for
agriculture on steep slopes.
Rank 3 (Cluster 7): Watersheds in this group have a score of
2, with poor scores for impoundment density and rate of
population change. Forests also tend to be fragmented into
small patches.
Rank 3 (Cluster 2): Watersheds in this group also have a
relative cumulative impact score of 2, with significant
population increases and forests that are no longer in
large, contiguous blocks.
Rank 4 (Cluster 1): This group has a score of 3, with poor
scores for population change, road density, and number of
impoundments per 1000 stream kilometers. Their score for
the amount of forest in large, contiguous blocks is in the
upper third.
Rank 5 (Cluster 3): Watersheds in this group have a relative
cumulative impact score of 4, with poor scores for the
amount of roads near streams, agriculture on steep slopes,
riparian cover, and sulfate deposition.
Rank 6 (Cluster 5): This group has a score of 5, with poor
scores for road density, agriculture on steep slopes, amount
of riparian vegetation and forest in large, contiguous
blocks, and sulfate deposition.
Rank 7 (Cluster 6): Watersheds in cluster 6 represent the
most urbanized areas in the region, including Pittsburgh,
Philadelphia, Washington, and Norfolk. With a relative
cumulative impact score of 6, only their values for
population change and agriculture on steep slopes are not
in the poorest third.
should, over time, result in an increase in the abundance of these species in these watersheds. Watersheds
in cluster 4 provide a good opportunity for reconnecting forests so that larger blocks are formed because
their overall percent forest cover is already high (Wickham and others 1999).
Summary and Conclusion
Many of today’s environmental problems can cooccur in space because they tend to be regional by nature.
Their cooccurrence requires that integrated assessments take account of cumulative impacts. We exam-
Integrated Environmental Assessment
ined the spatial pattern of some regional indicators to
develop relative cumulative impact scores for watersheds in the mid-Atlantic region. Relative cumulative
impact scores were highest around the major urban
centers, but, overall, there was no discernible gradient,
such as scores increasing from west to east.
Despite the exploratory nature of cluster analysis, it
appears to be a useful tool for grouping watersheds with
similar environmental characteristics. Nevertheless, the
data and methods used represent only a first attempt at
a regional-scale integrated environmental assessment.
More research is needed on other appropriate methods
and measurements.
559
Jones, K. B., K. H. Riitters, J. D. Wickham, R. D. Tankersley,
R. V. O’Neill, D. J. Chaloud, E. R. Smith, and A. C. Neale.
1997. An ecological assessment of the United States midAtlantic region: A landscape atlas. USEPA/600/R-97/130.
US Environmental Protection Agency, Office of Research
and Development, Washington, DC, 105 pp.
Kepner, W. G., K. B. Jones, D. J. Chaloud, J. D. Wickham, K. H.
Riitters, and R. V. O’Neill, R. V. 1995. Mid-Atlantic landscape indicators project plan—environmental monitoring
and assessment program. USEPA620/R-95/003. US Environmental Protection Agency, Office of Research and Development, Washington, DC, 37 pp.
Lee, L. C., and J. G. Gosselink. 1988. Cumulative impacts on
wetlands: Linking scientific assessments and regulatory
alternatives. Environmental Management 12(5):591–602.
Ligon, F. K., W. E. Dietrich, and W. J. Trush. 1995. Downstream
ecological effects of dams. Bioscience 45(3):183–192.
Acknowledgments
The authors thank Vern Meentemeyer, Ron Foresta,
Jim MacMahon, Kent Price, Jianguo Wu, Larry Harris,
and an anonymous reviewer for their many helpful
comments on earlier versions of the paper. The authors
also thank their collaborators at the USGS, EROS Data
Center for the continued commitment to innovative
development of geographic data.
Literature Cited
Aldenderfer, M. S., and R. K. Blashfield. 1984. Cluster analysis.
Sage University paper series on quantitative applications in
the social sciences, series no. 07-044, Sage Publications,
Beverly Hills, 88 pp.
Allen, T. F. H., and T. W. Hoekstra. 1987. Problems of scaling
in restoration ecology: a practical application. Pages 289–
299 in W. R. Jordan, M. E. Gilpin, and J. D. Aber (eds.),
Restoration ecology: a synthetic approach to ecological
research. Cambridge University Press, Cambridge.
Beanlands, G. E., W. J. Erckmann, G. H. Orians, J. O’Riordan,
D. Policansky, M. H. Sada, and B. Sadler. (eds.). Cumulative
environmental effects: A binational perspective. Canadian
Environmental Assessment Research Council/U.S. National
Research Council, Ottawa and Washington, 166 pp.
Environmental Management. 1988. Special issue on cumulative
environmental effects, Volume 12(5).
Evans, I. S., and K. Jones. 1981. Ratios and closed number
systems. Pages 123–134 in N. Wrigley and R. J. Bennett
(eds.) Quantitative geography: A British view. Routledge
and Kegan Paul, Boston, 419 pp.
Lynch, J. F., and D. F. Whigham. 1984. Effect of forest
fragmentation on breeding birds communities in Maryland,
USA. Biological Conservation 28:287–324.
Odum, W. E. 1982. Environmental degradation and the
tyranny of small decisions. Bioscience 32(9):728–729.
O’Neill, R. V., K. B. Jones, K. H. Riitters, J. D. Wickham, and I.
A. Goodman. 1994. Landscape monitoring and assessment
research plan—1994. EPA/620/R-94/009. US Environmental Protection Agency, Office of Research and Development,
Washington, DC, 53 pp.
O’Neill, R. V., C. T. Hunsaker, K. B. Jones, K. H. Riitters, J. D.
Wickham, P. M. Schwarz, I. A. Goodman, B. L. Jackson, and
W. S. Baillargeon. 1997. Monitoring environmental quality
at the landscape scale. Bioscience 47(3):513–519.
Peterjohn, W. T., and D. L. Correll. 1984. Nutrient dynamics in
an agricultural watershed: Observations on the role of a
riparian forest. Ecology 65(5):1466–1475.
Pielke, R. A., and R. Avissar. 1990. Influence of landscape
structure on local and regional climate. Landscape Ecology
4(2/3):133–155.
Preston, E. M., and B. L. Bedford. 1988. Evaluating cumulative
effects on wetland functions: A conceptual overview and
generic framework. Environmental Management 12(5):565–
583.
Rechow, A. K., M. N. Beaulac, and J. T. Simpson. 1980.
Modeling phosphorous loading and lake response under
uncertainty: A manual and compilation of export coefficients. EPA 440/5-80-011. US Environmental Protection
Agency, Washington, DC, 214 pp.
Fabos, J. G. 1985. Land use planning: From local to global
challenges. Chapman Hall, New York, 223 pp.
Renard, K. G., G. R. Foster, G. A. Weiss, D. K. McCool, and
D. C. Yoder (coordinators). 1997. Predicting soil erosion by
water: A guide to conservation planning with the revised
universal soil loss equation (RUSLE). US Department of
Agriculture, Agricultural Handbook No. 703, Washington,
DC, 404 pp.
Hunsaker, C. T., R. L. Graham, G. W. Suter, II, R. V. O’Neill,
L. W. Barnthouse, and R. H. Gardner. 1990. Assessing
ecological risk on a regional scale. Environmental Management 14(3):325–332.
Riitters, K. H., R. V. O’Neill, C. T. Hunsaker, J. D. Wickham,
D. H. Yankee, S. P. Timmins, K. B. Jones, K. B., and B. L.
Jackson. 1995. A factor analysis of landscape pattern and
structure metrics. Landscape Ecology 10(1):23–39.
Irwin, F., and B. Rhodes. 1992. Making decisions on cumulative environmental impacts: A conceptual framework. World
Wildlife Fund, Washington, DC, 54 pp.
Riitters, K. H., J. D. Wickham, and K. B. Jones. 1996. A
landscape atlas of the Chesapeake Bay watershed. 2nd ed.
Tennessee Valley Authority, Norris, 29 pp.
560
J. D. Wickham and others
Riitters, K. H., R. V. O’Neill, and K. B. Jones. 1997. Assessing
habitat suitability at multiple scales: a landscape-level approach. Biological Conservation 81:191–202.
Rykiel, E. J. 1985. Towards a definition of ecological disturbance. Australian Journal of Ecology 10:361–365.
SAS. 1989. SAS/STAT User’s Guide, Version 6, Vol. 1, 4th ed.
SAS Institute Inc., Cary, North Carolina, 943 pp.
Shoemaker, D. J. 1994. Cumulative environmental assessment.
Department of Geography Publication Series, Number 42,
University of Waterloo, Ontario, 129 pp.
Soranno, P. A., S. L. Hubler, S. R. Carpenter, and R. C.
Lathrop. 1996. Phosphorous loads to surface waters: a
simple model to account for spatial pattern of land use.
Ecological Applications 6(3):865–878.
USGS. 1982. Codes for the identification of hydrologic units in
the United States and the Caribbean outlying areas. USGS
Circular 878-A. US Geological Survey, Reston, Virginia.
USGS. 1989. Digital line graphs from 1:100,000-scale data.
Data user guide 2. US Geological Survey, National Mapping
Division, Reston, Virginia, 88 pp.
USGS. 1993. Digital elevation models. Data users guide 5. US
Geological Survey, National Mapping Division, Reston, Virginia, 48 pp.
Wessman, C. A. 1992. Spatial scales and global change:
bridging the gap from plots to GCM grid cells. Annual
Review of Ecology and Systematics 23:175–200.
Westman, W. E. 1977. How much are nature’s services worth?
Science 197:960–964.
Wharton, C. H., W. M. Kitchens, E. C. Pendelton, and T. W.
Sipe. 1982. The ecology of bottomland hardwood swamps of
the southeast: A community profile. FWS/OBS-81/37. US
Fish and Wildlife Service, Biological Services Program,
Washington, DC.
Wickham, J. D., J. Wu, and D. F. Bradford. 1997. A conceptual
framework for selecting and analyzing stressor data to study
species richness at large spatial scales. Environmental Management 21(2):247–257.
Wickham, J. D., K. B. Jones, K. H. Riitters, T. G. Wade, and R. V.
O’Neill. 1999. Transitions in forest fragmentation: Implications for restoration opportunities at regional scales. Landscape Ecology. 14(2):137–145.
Wischmeier, W. H., and D. D. Smith, D. D. 1978. Predicting
rainfall erosion losses—a guide to conservation planning.
US Department of Agriculture, Agriculture Handbook 537,
Washington, DC, 58 pp.
Xie, Y. 1995. The overlaid network algorithms for areal
interpolation problem. Computers, Environment, and Urban
Systems 19(4):287–306.